利用立体光流估计运动物体状态

Estimating the Moving Objects by Stereo Optical Flow

针对在无法获取相关视频内容先验知识的前提下进行视频中两帧之间物体运动预估的问题。提出利用双目立体标定结合立体光流场估计物体在三维空间中的运动的方法。介绍由时域光流和视域光流所组成的立体光流场的原理,推导利用立体光流场计算空间距离公式。通过实验数据进行分析表明该方法准确性较高,使机器基于光流法的双目视觉条件下对运动物体的判定有较好效果。

In order to obtain object motion estimation between two frames in the video for lack of prior knowledge of the video content, presents a motion estimation method （algorithm） in three-dimensional space based on binocular vision and stereo optical flow field. Firstly, proposes the principle of stereo optical flow field composed of time-domain optical flow and horizon optical flow. Then, deduces the calculation formula for space distance of stereo optical flow field. Finally, designs an experiment to confirm high accuracy and the effectiveness of the proposed method.